Past exam of the philosophy of science supplementary subject of the University of Oxford / 2019 (A14139W1)

Source paper: www.whatdotheyknow.com/request/oxford_physics_past_papers_2/response/2340696/attach/7/Physics%20Papers%202018.pdf?cookie_passthrough=1

The essays below address all eight questions, including both alternatives in questions 5 and 7. They defend particular reasoned positions rather than imply that the contested issues have unique model answers.

Thomas Kuhn's account is a useful corrective to a history in which scientists simply accumulate observations and replace theories whenever a contrary observation appears. It is less convincing as an exceptionless sequence governing every science. Its compatibility with scientific progress depends on what progress means.

In normal science, a community shares a scientific paradigm: exemplary achievements, accepted instruments and standards for choosing tractable problems. An unexpected result can initially be treated as a problem in the calculation or apparatus. This need not be irrational. A framework that succeeds elsewhere gives investigators reasons to check an instrument before abandoning the framework. Revolutionary theory change reorganizes some of those commitments instead of merely adding another measurement.^[ref]

The important contrast is between a descriptive account of how research is organized and a justification of every community's decision. Shared training can explain persistence without making persistence correct. A contrary experiment can expose a failed auxiliary hypothesis, but it can also expose a central error. The Duhem problem rules out an automatic inference from one anomalous reading to one uniquely false theory; it does not authorize indefinitely protecting a theory from independently calibrated tests.

Nor does incommensurability require complete inability to compare. Theories can change the meanings of such quantities as mass while still being compared through specified apparatus readings, limiting calculations and predictions. For instance, the relativistic energy of a free massive particle has the expansionAt $p\ll mc$, subtracting the rest energy leaves the familiar leading kinetic energy. This is an explicit connection between domains, not an assertion that every concept remained unchanged. It gives a concrete way to retain successful calculations while recognizing the earlier theory's limitations.

Progress measured as closer approach to a single final ontology is therefore not guaranteed by the historical pattern alone. Progress measured through more accurate prediction, solving independently specified problems and understanding why older models work in restricted domains is compatible with substantial conceptual change. These criteria must not be defined merely as whatever a victorious community happens to prefer.

Finally, different specializations can develop competing frameworks simultaneously. A research community need not occupy one uniform stage, and a chronological label alone does not demonstrate a crisis or revolution. The strongest use of Thomas Kuhn's account is consequently comparative: identify what changed in examples, standards and practice, then ask which changes improved scientific capacities. It explains important features of research without supplying a universal historical law or implying that all theory choices are equally good.

Some scientific classifications plausibly identify natural kinds; it does not follow that every useful theoretical category does. A good test asks whether the classification supports reliable inference because of relevant natural structure, rather than treating either practical usefulness or a precise verbal definition as sufficient.

A chemical element illustrates a strong case. Its atomic number fixes nuclear electric charge and constrains the electronic structure of the corresponding neutral atom. The classification thereby supports explanations of recurring chemical properties through quantum mechanics. Isotopes show why the claim must be qualified: equal atomic number does not imply equal mass, nuclear stability or every reaction rate. Whether a division is informative depends on which properties the investigation is trying to predict.

Biological classification need not have the same structure. The biological species concept emphasizes reproductive relations; a phylogenetic species concept emphasizes descent and diagnosable lineages. Hybridization, asexual reproduction and variation within a lineage make a single intrinsic essence an implausible universal criterion. A homeostatic property cluster theory instead asks whether interacting causal processes sustain a cluster of traits that supports induction, without requiring every individual to possess every trait.^[ref] These considerations can support realism about biological groupings without making all boundaries equally sharp.

There are also classifications whose role is expressly relative to a task. Calling a particle nonrelativistic compares its momentum scale with $mc$, or its speed with $c$. An investigator can set a tolerance for neglecting higher terms in an expansion. The boundary then depends partly on the accuracy required; it is not a newly discovered fundamental species of particle. Nevertheless, the approximation is constrained by measurable scales and can be objectively excellent or poor.

This prevents two opposite mistakes. Since a category is chosen for a purpose, it need not be arbitrary: a classification that repeatedly supports independent predictions has a better claim to natural grounding than an arbitrary grouping. Conversely, successful prediction does not uniquely identify the world's one privileged taxonomy. Different scales and explanatory questions may justify overlapping classifications of the same objects.

The defensible conclusion is selective and plural. Treat some kinds as tracking stable mechanisms, structures or historical lineages; assess the evidence for each classification and its intended domain. Retain useful idealized or conventional categories without pretending that every term in a scientific theory names an equally fundamental division of nature.

A Bayesian confirmation account can dissolve the apparent conflict between logical equivalence and ordinary judgements about relevant evidence. It does so by rejecting an unrestricted positive-instance rule and specifying how the evidence was sampled. Bayes' theorem alone does not guarantee that every white shoe is irrelevant, or that every black raven provides strong support.

Let $H$ assert that every raven is black. Its contrapositive is logically equivalent to $H$, so the two descriptions must receive the same conditional probability. The raven paradox arises if every positive instance of a universal statement automatically confirms it: a white nonraven then confirms the contrapositive and hence $H$. The disputed step is that unrestricted confirmation rule, not the logical equivalence.^[ref]

For background $K$, set $p=P(H\mid K)$, $a=P(E\mid H,K)$ and $b=P(E\mid \neg H,K)$. For $0<p<1$ and positive evidence probability,Thus it is the comparison of likelihoods, including the sampling protocol and alternatives, that determines support. Merely rewriting $H$ cannot change it.

The finite-population model of raven confirmation makes the intuitive difference quantitative. Fix $r$ ravens and $M$ nonblack objects. Compare an all-black-ravens hypothesis with the alternative that exactly one raven is nonblack. A uniform draw among nonblack objects that finds a nonraven has likelihood ratio $1/(1-1/M)$. A uniform draw among ravens that finds a black one has likelihood ratio $1/(1-1/r)$. If $M\gg r>1$, the former is much nearer one. With equal prior probabilities, $M=10^6$ and $r=100$ give posterior probabilities approximately $0.50000025$ and $0.50251256$. A tiny amount of confirmation need have no practical significance.

This example is not a theorem about every possible background. Inspecting an object already known to be a white shoe adds no relevant information about its being a nonraven. Other alternatives or selection mechanisms change the likelihoods and can remove or reverse the support. Conditioning only on the observed colour while silently changing how the object was selected would compare different experiments.^[ref]

The resolution is therefore qualified. A coherent probability model preserves equivalence and explains why superficially parallel observations need carry radically different evidence. It does not select a unique prior or sampling model without scientific judgement. Those remaining modelling questions are limitations of an application, rather than a contradiction in Bayesian inference.

Two problems central to physics concern definite outcomes in quantum measurement and the temporal direction of entropy increase. They expose different gaps between a dynamical description and the phenomena explained; neither can be resolved merely by saying that physics predicts observations successfully.

For the measurement problem, combine a two-state system with an apparatus initially in a ready state. If an ideal unitary operator correlates basis states with distinct pointer states, its linearity givesThe same dynamics that gives reliable correlations therefore gives an entangled final state, rather than a rule choosing one pointer value. The problem is not that the state has failed to predict possible outcomes, but how definite outcomes and their probabilities are related to that state.

Quantum decoherence explains why interaction with an environment suppresses interference in the apparatus's reduced density matrix: its off-diagonal term is multiplied by the inner product of the corresponding environmental states. Near-orthogonal records make this term very small. Tracing out the environment, however, does not by itself select one term in the global superposition. An objective-collapse theory changes the dynamics, while a many-worlds interpretation changes what is required of an outcome. Distinguishable modifications can be tested; interpretations reproducing the same predictions raise additional explanatory and conceptual questions. Decoherence solves a substantial dynamical part of the problem without automatically settling every interpretation.

The second issue is the thermodynamic arrow of time. In a classical model, time reversal in classical mechanics can pair an entropy-increasing trajectory with a reversed entropy-decreasing one. Liouville theorem in Hamiltonian mechanics preserves phase-space volume, so reversible Hamiltonian dynamics does not imply a universal monotonic increase of fine-grained entropy. The Second law of thermodynamics therefore needs a statistical and boundary-condition account, not a deduction from reversibility alone.

Boltzmann entropy assigns larger entropy to a macrostate occupying more compatible phase-space volume. Starting from a suitable low-entropy condition, overwhelmingly more accessible microstates can evolve toward larger macrostates. But an unrestricted typicality argument applied at an arbitrary low-entropy instant generally points toward increasing entropy in both temporal directions. The past hypothesis supplies an asymmetric low-entropy boundary condition in the past; one must still justify the probability measure and the macroscopic description used.

These are genuine partial resolutions rather than interchangeable cures. Decoherence concerns the accessibility of interference and records; a past boundary condition concerns temporal asymmetry in a statistical explanation. A satisfactory account must say which extra assumptions it introduces and which observable consequences follow. Neither problem licenses dismissing successful physics, but both show why a predictive formalism and a complete interpretation of its success are different achievements.

There are useful distinctions between science and pseudoscience, but one simple property of an isolated sentence is unlikely to classify every historical practice correctly. The relevant assessment should examine both empirical commitments and how a programme responds to evidence.

Falsifiability is a valuable starting point. A claim that excludes no possible observational outcome cannot receive the same kind of empirical support as one that risks failure. A prediction made before an experiment, with conditions and tolerances specified, is more informative than a story adjusted afterwards to fit any result. This distinguishes taking an empirical risk from merely finding a favourable description of what happened.^[ref]

However, falsifiability is neither sufficient for being good science nor identical to having evidence. An arbitrary precise prediction can be falsifiable and poorly motivated. A well-supported theory can also encounter an apparent failure caused by an auxiliary hypothesis about calibration or initial conditions. The Duhem problem explains why sensible investigators sometimes repair an apparatus model rather than immediately discard a theory. It does not justify repairs chosen solely to prevent every conceivable adverse result.

A more informative comparison asks whether revisions produce independently testable predictions, whether controls address plausible errors, whether unfavourable results are recorded and whether others can reproduce the procedures. An adjustment can be scientifically productive if it connects a former discrepancy with new phenomena. The same algebraic adjustment is much less impressive if it does nothing beyond reproducing the data used to choose it. Scientific reproducibility helps make a research practice corrigible, although reproducibility alone cannot establish that its interpretation is correct.

Historical judgement also requires separating an immature proposal from a practice organized to evade criticism. A new mechanism may initially lack a precise experimental test. Its current weakness is a reason to qualify confidence, not proof that no further development can make it scientific. Conversely, technical vocabulary and numerical output do not make a programme scientific when its central claims remain insulated from evidence.

The demarcation should therefore be explained through concrete features of inquiry: identifiable hypotheses, discriminating tests, calibration, treatment of alternatives and accountable revision. Different features can matter differently across observational, experimental and historical sciences. Borderline cases do not erase clear failures, just as the absence of a sharp temperature boundary between warm and cold does not make all temperatures equal.

This approach also keeps the conclusion proportionate. Calling a practice scientific does not guarantee truth, ethical acceptability or practical usefulness. Calling it pseudoscientific should identify the evidential or methodological failure, rather than replace the argument with a label.

Falsificationism captures an important asymmetry between universal claims and counterexamples, but it is an incomplete account of scientific appraisal. It works best as a demand for serious testing within fallible theoretical packages, rather than as a mechanical rule for discarding a theory after any anomalous observation.

For an unrestricted universal claim, finitely many successful cases do not deductively establish every unobserved instance. One genuine contrary case can contradict the universal claim, provided the predicates and circumstances are correctly identified. Karl Popper's emphasis on attempts at refutation consequently highlights something that mere accumulation of agreeable examples can miss: a test is informative partly because plausible alternative outcomes would have counted against the proposal.^[ref]

Real predictions nevertheless involve a conjunction. If $T\wedge A\Rightarrow E$ and $\neg E$, thenHere $A$ includes an apparatus model and initial conditions. The conclusion does not identify $T$ as the failed component. Independent calibration, varied experimental arrangements and agreement across instruments help distinguish the possibilities; they do not eliminate the logical structure of the Duhem problem. Insisting that every auxiliary assumption remain unrevisable would obstruct ordinary experimental repair.

Probabilistic theories create another difficulty for simple deductive falsification. If an ideal Bernoulli distribution model assigns a probability $0<q<1$ to success, every finite binary sequence has positive probability. An unusual sequence does not literally contradict the model. A statistical hypothesis test can instead use a predeclared rejection region with a controlled type I error. Its possible mistaken rejection is part of the method, not a contradiction to be ignored. Comparing alternatives also requires likelihoods or other explicit criteria, rather than a binary division into refuted and unrefuted proposals.

Finally, survival alone is not enough. A vague conjecture and a quantitatively successful theory might both be unrefuted; they need not be equally supported. Scientific decisions also concern approximation error, independent explanatory connections and whether a proposed modification predicts anything new. These considerations need not restore conclusive verification, but they require more than deductive refutation.

The plausible core is therefore methodological: specify what could count against a claim, expose it to demanding tests and resist changes made only to accommodate failure. The stronger thesis that this principle supplies a complete logic of discovery, theory choice and confirmation is not established. Critical testing remains valuable precisely when its assumptions and statistical limitations are stated honestly.

There are serious philosophical reasons to resist inferring the literal truth of every successful theoretical description. Those reasons support a selective and revisable attitude; they do not establish that none of the unobservable entities investigated by science exists.

Pessimistic induction starts from successful earlier theories whose central ontological claims were later rejected. Their history challenges an unrestricted inference from predictive success to truth. For example, successful descriptions of light transmission were once combined with an account involving a mechanical luminiferous aether. Continued use of some mathematical relations does not show that the proposed medium was real. The force of the objection depends on which part of a theory produced its successful predictions, not on collecting a list of old mistakes.^[ref]

Scientific underdetermination supplies a different challenge. Agreement with available data can leave rival accounts open. If they are empirically equivalent throughout a specified domain, another experiment restricted to that domain cannot discriminate them. This limits what those data establish. It does not show that every actual theoretical disagreement is permanently undecidable, or that equally compatible models must receive equal confidence given all relevant background evidence.

The realist reply emphasizes convergence across independent ways of detecting and manipulating a putative entity. An electron is not supported by one formula alone: accounts of its electric charge, scattering and response to fields connect different apparatuses and phenomena. Such success can be a reason to infer a common physical cause. The no-miracles argument expresses that explanatory motivation, but its force is abductive rather than deductive. It must not assume that only a literally true total theory could produce any success.

Several qualified positions remain defensible. Entity realism places particular weight on robust experimental use of entities without endorsing every theoretical account of them. Structural realism emphasizes specified relations retained across theory change. Neither label removes the need to identify what experimental interventions establish or which interpreted structure survives. Preserving an equation's shape alone is insufficient if the meanings and tested domains have changed.^[ref]

Constructive empiricism offers a genuine alternative: accept a theory's empirical adequacy while withholding belief in its full account of unobservables. The realist must explain why stronger belief is warranted, rather than accuse this stance of denying ordinary laboratory success.

A defensible realism therefore concerns particular well-supported claims, mechanisms and relations, with explicit domains and error allowances. It recognizes historical revision without treating every surviving term as equally secure. Neither universal literal realism nor universal scepticism about entities follows from scientific success or its historical failures.

Scientific underdetermination limits what evidence alone establishes, but its practical consequences depend on the strength of the equivalence being claimed. Rival theories fitting present data need not agree on future observations. Rival formulations agreeing on every observable in a specified domain present a more persistent limitation within that domain.

With finitely many data points $(x_i,y_i)$, suppose an interpolation polynomial $p$ fits all $n$ measurements. Thenalso fits them for every $\lambda$. This proves nonuniqueness for a deliberately broad model class. It does not prove that all these functions are equally plausible physical theories. Constraints from independently tested mechanisms, bounds on derivatives, symmetries and the measurement process can sharply restrict the class. A new observation at a different point can also distinguish the displayed alternatives.

The Duhem problem concerns an additional kind of latitude: a prediction tests a theory together with auxiliary hypotheses. A mismatch can be assigned to different components. This is a reason to vary apparatus, calibrate independently and use multiple kinds of evidence. It is not a demonstration that any preferred theory can be made equally well supported by arbitrary repairs.^[ref]

Exact empirical equivalence requires more. If two accounts supply identical probability distributions for all observations allowed by a stated setup, observations of that kind cannot favour one through their likelihoods. Bayesian inference makes the limitation explicit: posterior odds equal prior odds when the likelihood ratio is one. Declaring a prior preference does not transform it into new experimental discrimination.

Theories can nevertheless differ in explanatory organization, computational usefulness or how they connect with a larger domain. Those are legitimate questions, provided their role is stated. If a proposed new intervention distinguishes them, the equivalence did not cover that enlarged domain. If the difference is instead only a change of redundant description, treating it as two competing physical worlds may create a spurious dispute; gauge symmetry provides an important case requiring care.

For scientific practice, the appropriate response is model comparison with explicit assumptions, searches for discriminating tests and honest limits on what has been established. For scientific realism, enduring empirical equivalence challenges any claim that the evidence uniquely fixes a full ontology. It does not make precise predictions or experimental decisions impossible. Underdetermination is consequently a constraint on inference whose seriousness must be demonstrated case by case, rather than a general proof that all scientific reasoning is arbitrary.

A commitment to testing claims against experience is a defensible scientific norm. A strict account according to which every meaningful scientific statement must be conclusively verified, or translated into a fixed vocabulary of theory-free observations, is much harder to sustain. Logical empiricism should also be understood as a changing family of projects, not reduced without qualification to its strongest early formulations.

An unrestricted universal scientific law concerns cases beyond a finite observation record. No finite list of favourable cases deductively entails all of them. Requiring conclusive verification would therefore exclude much ordinary theoretical science. Replacing that demand with scientific confirmation retains a relation to evidence while accepting that support is fallible. This is a substantive revision, rather than proof that universal laws were conclusively verified after all.^[ref]

A second difficulty concerns observations. A detector count relies on assumptions about sensitivity, background and the production of a signal. Theory-ladenness makes a sharp division between completely neutral observations and theoretical claims implausible. But dependence on some assumptions does not imply dependence on the truth of the target theory. A calibration can be supported independently, and agreement across instruments using different mechanisms can make a putative effect less vulnerable to one shared error.

Nor should unobservable vocabulary be dismissed merely because its referents cannot be seen directly. A theory of electrons and electromagnetic fields can organize measured currents, spectra and scattering through a common model. What matters scientifically is how the theoretical commitments constrain observations and survive attempts to discriminate them from alternatives. Insisting that each term independently translate into an observation report can miss the role of a whole theoretical structure.

An empiricist can accept much of this. Constructive empiricism permits literal theoretical discourse and commitment to using a theory while restricting belief to empirical adequacy. It separates withholding a stronger ontological belief from refusing to reason with theoretical models.^[ref] Whether this restraint is preferable to selective scientific realism remains an argument about what successful testing warrants.

The strongest sustainable empiricism is consequently methodological and fallible: demand explicit empirical consequences, scrutinize calibration and preserve the possibility of correction. The stricter verificationist reduction does not adequately describe how laws, theoretical terms and measurement actually work. Rejecting that reduction need not weaken the central demand that scientific claims remain accountable to evidence.

A philosophical account of scientific explanation must accommodate causal explanations, but there is no good reason to insist that every explanation be causal or that every one explicitly contain an exceptionless universal law. The issue is what explanatory task an account performs, rather than whether it fits one syntactic template.

The deductive-nomological model explains by deriving a phenomenon from scientific laws and specified circumstances. It illuminates a real feature of physical reasoning: a result is intelligible because it follows from a more general structure, not merely because it has been listed. Yet valid derivability does not automatically supply the direction or relevance of explanation.^[ref]

For a vertical pole of height $h$ in a simple geometrical model, a solar elevation $\alpha$ gives a shadow length $s=h\cot \alpha$. Algebra permits inference from $h,\alpha$ to $s$, or from $s,\alpha$ to $h$. The equation alone does not decide which quantities are causal inputs. A causal explanation must supplement the relation with how the pole, illumination and surface generate the shadow, and with the changes meant by an intervention. Mere reversal of a calculation is not reversal of a physical mechanism.

Conversely, some explanations appeal to constraints rather than a mechanism producing an event. Twenty-three objects cannot be divided into three equal groups of whole objects because $3$ does not divide $23$. The relevant divisibility fact explains an impossibility without describing a cause that prevents each attempted arrangement. Physical applicability still requires that the objects count as indivisible for the task; the arithmetic alone is not a complete theory of the material.

Statistical explanation supplies another limit to a universal deductive requirement. A radioactive decay model gives a hazard rate and the probability $1-e^{-\lambda t}$ of decay within time $t$. It can explain rates and the distribution of waiting times without deductively fixing a particular atom's decay instant. A very unlikely event can still have a recognized mechanism. One must distinguish explaining why an event was possible or had a certain chance from explaining why this event occurred at precisely this time.

A demand for explicitly stated fundamental laws can also be too restrictive in a local mechanistic explanation. Identifying a broken circuit connection and the resulting interruption of current may answer a practical question without deriving the fault from microscopic dynamics. The explanation depends on physical regularities and background assumptions, but dependence is weaker than requiring every law to be displayed in the explanatory argument.

A useful plural account therefore asks whether the explanation identifies relevant dependence, states its domain and answers the actual contrastive question. Causal explanation, mathematical constraints, statistical accounts and explanatory unification can each contribute. This is not permission for arbitrary stories: relevance, independent support and the distinction between prediction and explanation remain constraints on all of them.